Today we announce the baseline accuracy for the fake news challenge and provide code which allows the public to reproduce and build-upon the features/classifier to improve the scores on the hold-out test set. Target to beat: 79.52% Learn more about the Fake News Challenge: http://fakenewschallenge.org Code is available on GitHub: https://github.com/FakeNewsChallenge/fnc-1-baseline
I’ve been awarded a £500 fund from the University of Sheffield Learned Society to support travel to EACL2017 in Valencia. Thank you.
Our software demonstration showing our recent work on verifying numerical claims has been accepted for the European chapter of the Association for Computational Linguistics 2017 conference in Valencia. Look forward to being able to going out there to demonstrate this work ✈️
Ok. A fully automated system for HeroX was harder that I thought it would be. But it’s highlighted some interesting challenges and having our ideas validated by journalists on real-world claims has been very helpful. We came third place (out of three). But we took on an additional challenge of being fully automated in our […]
I’d like to thank Microsoft for awarding a $5000 grant to support my ongoing research into fact checking. This will be used to purchase Bing queries to collect web pages that form part of my dataset for this project.
I’m entering the HeroX fact checking challenge with Andreas on behalf of my department. We’ll be developing a fully automated fact checking system that is able to link claims made in natural language to entries in a structured knowledge base and assign a truth label.
I’ve just started my PhD in Natural Language Processing at Sheffield University. I’ll be working with Dr Andreas Vlachos, researching how AI can be used to fact check information. Looking forward to the challenges ahead!